Determinant of flexible Parametric Estimation of Mixture Cure Fraction Model: An Application of Gastric cancer Data

Cure models are survival models basically developed to estimate the proportion of patients cured in a clinical trial. These models estimate the cured proportion and also the probability of survival. Cure models are a special type of survival analysis model where it is assumed that there are a proportion of subjects who will never experience the event and thus the survival curve will eventually reach a plateau. Cure has become an important measure of long term survival benefit derived from therapy. This study was intended at determining the flexible Parametric Cure Fraction Model for Gastric cancer Data. Suitability of four parametric mixture cure models were considered namely; Log Normal (LN) cure fraction model, Log Logistic (LL) cure fraction model, Weibull (W) cure fraction model and Generalized Gamma (GG) cure fraction model. AIC, mean time to cure), variance and cure fraction (c) were used to determine the flexible Parametric Cure Fraction Model among the considered models. Gastric Cancer data from 76 patients received adjuvant CRT and 125 receiving resection (surgery) alone were used to confirm the suitability of the models. The data was from a retrospective study in patients with gastric adenocarcinoma who underwent curative resection with D2 lymphadenectomy in the Barretos Cancer Hospital between January 2002 and December 2007. The survival for this data refers to the times until death in months since surgery.The Log-Logistic (LL) gave the minimum value for AIC, minimum means, minimum mean time to cure ( ) as well as cure fraction with values (525.865, 272.671); (0.529, 0.583); (1.697, 1.791) and (0.122, 0.123) using GCS and GCC, respectively. Also, GG gave the highest cure fraction with C = 0.374 and 0.599 for both GCS and GCC. Cure fraction obtained using GG presented the model as being best in terms of proportion to cure in GCS and GCC. The LogLogistic (LL) showed a promising result being the best flexible model in terms of AIC, minimum variance, means and mean time to cure, while Generalized Gamma performed best in terms the value of cure fraction which gives the proportion of cured.